201119414 六、發明說明： 【發明所屬之技術領域】 本發明係指一種影像插補處理裝置，尤指一種用以決 定一插補晝面中影像覆蓋區/影像顯露區之影像的影像處 理裝置及其相關方法。 【先前技術】 目前傳統影像插補機制在決定一插補晝面中之一插補 區塊的移動向量時，係直接以區塊比對演算法（block matching algorithm)的運算結果來決定其移動向量，來產 生該插補區塊的影像。請參考第1圖，其係為區塊比對演 算法的示意圖。如第1圖所示，影像畫面Fn-1、Fn係為輸 入影像中前、後兩張的晝面，而插補晝面Η係為傳統影像 插補機制所產生的一插補晝面。其中，區域R4及區域R5 係代表影像晝面Fn-1、Fn中的前景物體（Foreground Object)，其餘部分為背景影像，且圖中的第—參考移動向 量MV1表示一前景物體移動的方向，而第二參考移動向量 MV2則表示一背景影像的移動方向。此外，影像晝面 中區域R3’在影像晝面Fn-1中被前景物體遮住，故未出現 影像畫面Fn-Ι中；相同地’影像畫面Fn-Ι中區域R6，在影 像晝面Fn中被前景物體遮住，故未出現影像晝面中汾 因此，使用傳統的區塊比對演算法估算移動向量， : 月 &、 後兩張影像畫面Fn-1、Fn中皆可找到背景影像中區域幻 201119414 j^2 ^ R.7 τΕ»-» 決定出和处，以及前景物體中區域R4和R5,所以，可 κΆτΤ目標移動向量，並將影像區域幻、R2、R4、 和R8正確地呈現於插補晝面FI上（如第丨圖所 示）。 u厂, 動向ί日*，在決定插補晝面Η申區域R3與R6的目標移 -Fnm’由於插補晝面π中區域R3僅出現在影像畫面 斤括μΓ/幻中，未出現在影像晝面Fn_1中，所以在進行 — 區鬼較時，無法在前、後兩張影像晝面Fn-卜Fn中找到 • ==或，並用以決定移動向量;相同地，插補晝面Π °° 6僅出現在影像晝面Fn-Ι的區域R6,中，夫屮转| &中’所以在進行區塊比較時’無法在前、後 動向^。^如·卜如中卿目同的區域，並用以決定移 使用傳統的區塊比對演算法估算移動向量， =域一6的因素’會造成插補晝面η的整;品 2上’插補晝面FI中區域R3與R6内的移 • 應為“影像的移動向量，以使得區域R3與妨 里 ： ^景影像’例如，區域R3應使用第二參考移動/f ⑽R6亦應使用第二參考移動向量㈣201119414 VI. Description of the Invention: [Technical Field] The present invention relates to an image interpolation processing device, and more particularly to an image processing device for determining an image of an image coverage area/image display area in an interpolation surface and Its related methods. [Prior Art] At present, when the conventional image interpolation mechanism determines the motion vector of one of the interpolation blocks, the motion of the block matching algorithm is directly determined by the operation result of the block matching algorithm. Vector to generate an image of the interpolation block. Please refer to Figure 1, which is a schematic diagram of the block alignment algorithm. As shown in Fig. 1, the image frames Fn-1 and Fn are the front and back faces of the input image, and the interpolation face is an interpolation face created by the conventional image interpolation mechanism. Wherein, the region R4 and the region R5 represent foreground objects in the image planes Fn-1 and Fn, and the rest are background images, and the first reference motion vector MV1 in the figure indicates the direction in which the foreground object moves. The second reference motion vector MV2 represents the moving direction of a background image. In addition, the region R3' in the image plane is covered by the foreground object in the image plane Fn-1, so that the image frame Fn-Ι does not appear; the same image screen Fn-Ι region R6, in the image plane Fn The image is hidden by the foreground object, so there is no image in the face. Therefore, the traditional block comparison algorithm is used to estimate the motion vector: Month &, the last two image frames Fn-1, Fn can find the background The area in the image is 2011201114 j^2 ^ R.7 τΕ»-» Determines the sum and the areas R4 and R5 in the foreground object, so the κΆτΤ target moves the vector and the image area is illusory, R2, R4, and R8 Appear correctly on the interpolated face FI (as shown in the figure below). u factory, movement ί 日*, in the decision to interpolate the surface of the region R3 and R6 target shift -Fnm' due to the interpolation surface π in the region R3 only appears in the image frame, not appearing in the image The image is in Fn_1, so it is not possible to find • == or in the front and back two images Fn-b Fn when performing the time zone ghost; and use it to determine the motion vector; ° ° 6 only appears in the area R6 of the image plane Fn-Ι, in the middle of the | 屮 & & & & & & & 所以 所以 所以 所以 所以 所以 所以 所以 所以 所以 所以 所以 所以 所以 所以 所以 所以 所以 所以 所以^ Such as Bu Ruzhong, the same area, and used to determine the use of traditional block comparison algorithm to estimate the motion vector, = domain 6 factor 'will cause the interpolation of the face η whole; The interpolation in the areas R3 and R6 in the interpolating FI plane should be "moving vector of the image so that the area R3 and the scam: ^ scene image", for example, the area R3 should use the second reference movement /f (10) R6 should also be used Second reference motion vector (4)
Fn-1的區域R6，中找到對應的晝面區域。所以；：面 :算二並無:決定出正確的移動向量，造成插補區域= w王現影像失真的狀況，當應用於圖框率轉換： 她⑽簡⑽時’傳統影像插補機制將會^ 影像的品質。 田尸牛低輪出 201119414 【發明内容】 因此，本發明的目的之一在於提供一種能夠正確地決 定-插補區塊之影像的影像處理裝置及其相關方法，以解 決上述的問題。 依據本發明的實施例，其係揭露一種影像插補處理農 置。該影像插補處理裝置，用於進行運動畫面插補時，決 定一插補晝面中影像覆蓋區及影像顯露區之一代表性移動 向置，該插補晝面由複數個區塊所組成，該裝置包含有移 ，向里產生模組，用於產生該插補畫面中—插補區塊之— 第=參考移動向量與-第二參考移動向量；—模糊區塊處 ，模組’肖於依據該桐肖區塊之該參考移動向量與該 第一參考移動向量來決定該插補區塊之一區域屬性係為影 像覆蓋區及影像顯露區的其中之-，及該插補區塊之一= 表，移動向置係為該第一參考移動向量與該第二參考移動 向置的其巾之―；以及—晝面插補模組’祕依據該代表 / 生移動向昼及該區域屬性，來產生該插補區塊。 、、，據本發明的實施例，其另揭露一種影像插補處理方 二亥办像插補處理方法，用於進行運動晝面插補時，決 ^插補晝面中影像覆蓋區及影像顯露區之—代表性移= 向里°亥插補晝面由複數個區塊所組成，該方法包含有產 生插補晝面中-插補(I塊之-第-參考移動向量與一第 一麥=移動向量；依據該插補區塊之該第一參考移動向量 亥第一參考移動向量來決定該插補區塊之一區域屬性係 201119414 為影像覆蓋區及影像顯露區的其中之一，及該插補區塊之 代表性移動向量係為該第一參考移動向量與該第二參考 移動向量的其中之-；以及依據該代表性移動向量及該區 域屬性，來產生該插補區塊。 本餐明之貫加例的優勢在於’若一區塊實際屬於一影 - 像覆蓋區或一影像顯露區，則上述影像插補處理裝置及方 ： 奸決找區塊之背景移動向量U-參考移動向量及第 =參考移動向量之其中之及區域屬性（影像覆蓋區及 • *像顯露區之其中之一）’來產生正確的影像，因而可提升 影像品質；而若該區塊實際並非屬於影像覆蓋區或影像顯 露區，則該影像插補處理裝置及方法可參考該區塊之原始 移動向量來產生正確的影像，因而不會影響影像品質 中該原始移動向量係以區塊比對演算法針對親塊所計^ 的一移動向量。 ^ 【實施方式】 ' 、、首先，為方便閱讀，以下係將前一張影像晝面中並未 ： 被前景物體遮住但卻在下一張影像晝面中被前景物體所遮 住^背景影像區域，稱作影像覆蓋區（c〇vered _)，並 將則一張影像晝面中被前景物體遮住但卻在下一張影像壹 2中出現（未被前景物體所遮住）的背景影像區域^稱= 影像顯露區（uncovered area);舉例來說，第i圖所示之 插補晝面Π中區域R6為影像覆蓋區，而區域幻為影像 頭路區；請注意，以上定義僅用以方便說明本發明之實施 201119414 例的操作，並非本發明的限制。 理裳第2圖’其繪示本發明一較佳實施例之影像處 :置ίο之功能方塊圖。影像處理裝置价制於進行運 時決定-插補畫面之影像覆蓋區/影像顯露區 向I’其中該插補晝面由複數個區塊所組成。如第 圖所示’影像處理裝i 10包含有一儲存單元刚及 單元·’其中儲存單元刚係用於儲存複數張原 。息面、插補晝面及使用(1塊比對產生之原始移動向量； 而=動插補單元200耦接至儲存單元1〇〇，包含有—移動 向1產生模組210、一模糊區塊處理模組22〇盥一查 :=組230，其中移動向量產生模組⑽係用來“該插 =面（亦即第—插補畫面）中之—區塊的第—參考移動 。置，VI與第二參考移動向量體。在本實施例中，移 白里產生模組210係對該插補晝面中的每一區塊皆產生 其=對應的兩個參考移動向量Μν^Μν2，每一區塊例如 已3有16x8的像素範圍，然此並非本發明的限制，1他像 素範圍的實施變型亦符合本發明的精神。接著，模糊區塊 處理拉組220係包括一模糊區塊判定模組221和一覆蓋/ 顯露區塊判定及移動向量決定模組222，係依據該區塊之 第一參考移動向量MV1與第二參考移動向量MV2來決定 該區塊位於影像覆蓋區或影像顯露區，並決定該區塊之一 代表性移動向量。其ψ，模糊區塊判定模組221依據該區 塊之第一、第二參考移動向量MV1與MV2的差異性決定 。亥區塊疋否位於模糊區域。而覆蓋/顯露區塊判定及移動向 置決定模，组222貝情位於模糊區域之區塊進行覆蓋/顯露區 201119414 塊判疋及代表性移動向量決定。最後，晝面插補模組23〇 會依據該代表性移動向量及該區塊位於影像覆蓋區或影像 顯露結果’來產生該區塊的插補畫面影像。 μ IV、上在本發明之實施例中，移動插補單元2⑽至 夕於原，晝面Fn_2、Fn]、如和Fn+1間各產生一插補晝 . 2以第4A〜4D圖的原始晝面Fn 2、如卜如、Fn+i等 影像為例來說’前景顧的㈣方向係由左至右水平移 動，而月景影像係由右至左水平移動；需注意的是，在此 φ $方便4明係僅以水平移動的例子作為說明，然而，對於 々特疋方肖（例如垂直方向或其他4壬何角度的移動方向 等）的影像移動’皆可使用本發明之實施例的影像處理裝 置進行處理。 以下先請參照第3A、3B圖係為一插補區塊產生兩參 2向里MV卜MV2之示意圖。移動向量產生模組21〇 母插補晝面之每一插補區塊產生兩參考移動向In the region R6 of Fn-1, the corresponding kneading region is found. So;: face: no two: determine the correct motion vector, resulting in the interpolation area = w king image distortion, when applied to the frame rate conversion: she (10) simple (10) 'traditional image interpolation mechanism will Will ^ the quality of the image. FIELD OF THE INVENTION Accordingly, it is an object of the present invention to provide an image processing apparatus and related method capable of accurately determining an image of an interpolation block to solve the above problems. In accordance with an embodiment of the present invention, an image interpolation process is disclosed. The image interpolation processing device is configured to determine a representative moving orientation of one of the image coverage area and the image display area in the interpolation plane when the motion picture interpolation is performed, and the interpolation surface is composed of a plurality of blocks. The device includes a shifting, inward generating module for generating the interpolation screen - the interpolation block - the first reference movement vector and the second reference movement vector; - the fuzzy block, the module ' According to the reference motion vector of the Tongxiao block and the first reference motion vector, it is determined that one of the region attributes of the interpolation block is the image coverage area and the image exposure area, and the interpolation area One of the blocks = the table, the moving orientation is the first reference movement vector and the second reference moving the orientation of the towel; and - the face interpolation module 'secret according to the representative / student movement The region attribute is used to generate the interpolation block. According to an embodiment of the present invention, a method for interpolating and processing an image interpolation processing method for performing image interpolation and clipping of the image coverage area and image in the interpolation plane is further disclosed. The exposed area - representative shift = the inward ° Hai interpolated plane consists of a plurality of blocks, the method includes the generation of interpolation - interpolation (I block - the - reference movement vector and a a MG=moving vector; determining, according to the first reference motion vector of the interpolation block, a first reference motion vector to determine one of the interpolation block region attribute system 201119414 as one of an image coverage area and an image exposure area And a representative motion vector of the interpolation block is one of the first reference motion vector and the second reference motion vector; and generating the interpolation region according to the representative motion vector and the region attribute The advantage of this meal is that if a block actually belongs to a shadow-like coverage area or an image-exposed area, then the image interpolation processing device and the side: the background movement vector U of the traitor - reference motion vector and reference = The motion vector and the region attribute (image coverage area and * * one of the exposure areas) are used to generate the correct image, thereby improving the image quality; and if the block does not actually belong to the image coverage area or the image is revealed The image interpolation processing device and method can refer to the original motion vector of the block to generate a correct image, and thus does not affect the image quality. The original motion vector is calculated by the block comparison algorithm for the parent block. A moving vector of ^. ^ [Embodiment] ', First, for the convenience of reading, the following image will not be in the previous image: it is covered by the foreground object but is foreground object in the next image The area of the background image is called the image coverage area (c〇vered _), and the image is covered by the foreground object but appears in the next image 壹2 (not by the foreground object) The background image area of the mask is = uncovered area; for example, the area R6 in the interpolated area shown in Fig. i is the image coverage area, and the area is the image head area; please It is to be noted that the above definitions are merely for convenience of explaining the operation of the embodiment of the present invention, which is not limited to the present invention. FIG. 2 is a functional block diagram of the image of the preferred embodiment of the present invention. The image processing device is priced at the time of operation determination - the image coverage area/image display area of the interpolation screen is I', wherein the interpolation surface is composed of a plurality of blocks. As shown in the figure, 'image processing equipment i 10 includes a storage unit and a unit · 'the storage unit is just used to store a plurality of originals. The surface, the interpolation surface and the use (the original movement vector generated by one alignment; and the = dynamic interpolation unit 200 coupled Connected to the storage unit 1A, including a moving to 1 generating module 210, a fuzzy block processing module 22, a check: = group 230, wherein the motion vector generating module (10) is used to "the plug = In the face (ie, the first interpolation screen) - the first reference movement of the block. Set, VI and the second reference move vector body. In this embodiment, the shift generation module 210 generates two reference motion vectors Μν^Μν2 corresponding to each of the blocks in the interpolation plane, and each block has, for example, 3 has 16×8 The pixel range is not limited by the present invention, and an implementation variant of the pixel range is also in accordance with the spirit of the present invention. Next, the fuzzy block processing pull group 220 includes a fuzzy block determining module 221 and a coverage/exposure block determining and moving vector determining module 222 according to the first reference moving vector MV1 and the second of the block. The motion vector MV2 is referenced to determine whether the block is located in the image coverage area or the image exposure area, and a representative motion vector of the block is determined. Thereafter, the fuzzy block decision module 221 determines the difference between the first and second reference motion vectors MV1 and MV2 of the block. The block in the block is located in the blurred area. The coverage/exposure block decision and the moving direction decision mode, the group 222 is located in the block of the fuzzy area for coverage/exposure area 201119414 block judgment and representative motion vector decision. Finally, the face interpolation module 23 产生 generates an interpolation picture image of the block according to the representative motion vector and the block located in the image coverage area or the image exposure result ′. μ IV, in the embodiment of the present invention, the mobile interpolation unit 2 (10) to the original, the face Fn_2, Fn], such as and Fn+1 each generate an interpolation 昼. 2 to 4A~4D The original image Fn 2, such as Bu Ru, Fn+i, etc., for example, the foreground (four) direction moves horizontally from left to right, while the moon image moves horizontally from right to left; Here, the φ $ convenience 4 is only explained by the example of horizontal movement, however, the invention can be used for the image movement of the 疋 疋 ( (for example, the vertical direction or the movement direction of other angles, etc.) The image processing apparatus of the embodiment performs processing. In the following, please refer to the 3A and 3B diagrams as an interpolation block to generate a schematic diagram of two parameters 2 inward MV MV2. Each interpolation block of the motion vector generation module 21 母 parent interpolation surface generates two reference movement directions
的運作來朗。首先，義向量產生模組⑽ “區塊比對演算料算相雜觀塊的移動向量，再 算出的移動向量與其複數個鄰近區塊的複數 個移動向1 ’來計算鶴向量的混_度，以算出一數值 本實施例來說，該數值曲線係為沿某—插補區塊 向量混亂程度所形成的曲線，而該混乱程 又、〇 里的變異值（motion vector variance)，亦即， 向上代表不同移動向量變 j又的複數個數值。請參照第μ圖，其 代表移動向量變異程度之數值㈣^= 201119414 補區塊與其複數個鄰近區塊（如第3a圖所示，位於5x5 的區塊範圍）依區塊比對演算法所計算出的複數個移動向 量刀別疋MVqo與MV·2·2〜MV22 ’則根據該等移動向量可 計算出一移動向量變異值MV_VAR，其計算方式是取該等 移動向量中最大水平分量減去其中最小水平分量所得的絕 對值，再加上該等移動向量中最大垂直分量減去其中最小 垂直分量所得的絕對值，MV_VAR可利用以下的等式表示 之： MV_VAR= | MAX(MVx)-MIN(MVx) | + I MAX(MVy)-MIN(MVy) | 等式(1) 其中MVX和MVy分別代表水平分量（x轴分量）與垂直分 量（y軸分量）。需注意的是，5x5的區塊範圍並非是本發 明的限制，其亦可利用NxN或是ΝχΜ的區塊範圍來實作 之，其中參數Ν與Μ皆是正整數且Ν不等於Μ ;此外， 計算移動向量變異值MV_VAR的方式亦可改用等式(2)或 等式(3)來實現： MV_VAR= I MAX(MVx)-MIN(MVx) | + I MAX(MVy)-MIN(MVy) | +SAD 等式(2) MV_VAR=ax{ | MAX(MVx)-MIN(MVx) | + I MAX(MVy)-MIN(MVy) | }+y3xSAD 等式(3) 其中數值SAD係為第一插補區塊依照區塊比對演算法所 异出的一區塊比對差值，參數a、0係為加權參數；凡可 用以計算代表移動向量變異程度之數值的任一實施變化， 皆屬於本發明的範疇。由上所述，依據等式(丨）、等式(2) 或等式(3)其中之一，移動向量產生模組210逐一針對不同 201119414 的插補區塊進行計算，如此可得出第一數值曲線cv ,如 第3B圖所示。 第3B圖所繪示為是本發明包含有多個移動向量變異 值之第一數值曲線CV«MB(K)係為第一插補區塊，而在第 3B圖中移動向量產生模組21〇係決定出插補區塊mb⑻的 兩參考移動向量MV：l、MV2,首先，以背景或前景影像移 動的方向（例如水平方向）來看，移動向量產生模組2⑺ ' 在空間上延著第一插補區塊MB00的兩侧複數個（例如 MB〇0水平兩侧各六個相同大小的區塊，mb10〜MB6〇與 ΜΒ·1()〜MB_6〇)區塊的相對應移動向量變異值，在該等相 對應移動向量變異值範圍内取出一極大值（例如第3B圖 所示之VARm ax )’並在該等相對應移動向量變異值範圍内 找出該極大值VARmax左、右兩側之極小值所對應到的區 塊’例如，可找到區塊娜-4〇與MB5〇，而此左、右兩區塊 MB-4〇與MBs。分別利用區塊比對演算法所計算之移動向量 即作為第一插補區塊MB00的兩參考移動向量MV1、 • MV2 ’換句話說’第-插補區塊MB。。的第一參考移動向 fMV1係對應至位於帛—數值輯CV之極大值vARmax 之左側的極小值VARmin，而其第二參考移動向量猜2係 ' 職至位於第—數值曲線CV之極大值VARmax之右側的 極小值VARmin’ ；第-、第二參考移動向量應、應2 的其中之-係對應至背景移動向量（backgr〇und — _〇Γ )，而其另一則對應於前景移動向量（foreground motion vector)，這是因為屬於影像覆蓋區或影像顯露區之 插補區塊週圍的移動向量變異值將會相當大，而左、右兩 201119414 侧之最小移動向量變異值所對應之影像.區塊，其會對應於 一前景移麟量或-背景移畅量，視其位於影像覆蓋區 或影像，露區而定。因此’若第一插補區塊mb⑽係位於 影像覆蓋區與影像顯露區之其中之一，則其第一、第二夫 考移動向量顺、體的其中之一對應於背景移動向 ，其另-對應於前景移動向量。藉由上述的操作，移動向 置產生模組210可計算出對插補晝面π中每一插補區塊的 相對應第一、第二參考移動向量^丨與_2。 接著，請參照帛4A、4B、4C及4D圖係為決定一模 糊區塊位於影像覆魏或影像顯露區及其代表性移動向量 之第了實施例示意圖。模糊區塊判定模組221係依據該區 塊之第一、第二麥考移動向量MV1與MV2的差異性決定 該區塊疋否位於模糊區域。當該區塊的第一、第二彖考移 動向量讀與MV2相似時，則該區塊不位於模糊區域^ 然當該區塊的第-、第二參考移動向量顺與Mv2差異 大時，則該區塊位於模糊區域。差異值“乂一^^汗的計算方 式是取參考移動向量MV1的水平分量減去第二參考 移動向量MV2水平分量所得的絕對值，再加上第—參考移 動向量MV1的垂直分量減去第二參考移動向量MV2的垂 直为1所得的絕對值’ MVjDiff可利用以下的等式表示之： MV_Diff= | MV1x-MV2x I + I MVly-MV2y | 等式(4) 上述等式(4)僅為本案—較佳實施方式，並非本發明的限 〇 接著，覆蓋/顯露區塊判定及移動向量決定模組222進 步對位於模糊區域之區塊進行覆蓋/顯露區塊判定及代 12 _ I _ 201119414 表性移動向量決定。在第4A圖中，插補晝面FI由原始畫 面Fn-Ι和Fn所插補產生’其中區域R3和區域R6中的區 塊位於模糊區域。以下以區域R3為例，詳細說明本實施 例作法。一待決定區塊Ο位於區域R3中，區塊Ο於原始 晝面Fn-Ι中被前景物體所遮蔽，未出現在原始晝面Fn-1 中，但有出現在原始晝面Fn中（意即區域R3位於影像顯 ： 露區），且區塊0的兩移動向量分別為第一、第二參考移 : 動向量MV1與MV2。首先，參照第4A圖，利用第二參 考移動向量MV2由區塊Ο對應至原始畫面Fn-Ι的區塊B • 及原始晝面Fn的區塊A ;接著，參照第4B圖，利用第一 參考移動向量MV1由區塊A及區塊B向前分別對應至區 塊A’及區塊B’。更詳細地說，原始晝面Fn中的區塊A向 前對應至原始晝面Fn-Ι中的區塊A’及原始晝面Fn-Ι中的 區塊B對應至原始晝面Fn-2的區塊B’。最後，本較佳實 施例採用區塊比對絕對差異總值（Sum of Absolute Difference)的方式計算區塊A和區塊A’的差異值SADI 及區塊B和區塊B’的差異值SAD2。同樣地，再利用第一 參考移動向量MV1分別由區塊A及區塊B向後對應至晝 面Fii+1的區塊A”及晝面Fn的區塊B ’ ’ ’並計算產生區塊 ' A和區塊A”的差異值SAD3及區塊B和區塊B”的差異值 SAD4。 然後，參照第4C圖，利用第一參考移動向量MV1由 區塊0對應至原始晝面Fn的區塊C及原始晝面Fn-Ι的區 塊D ;接著，參照第4D圖，利用第二參考移動向量MV2 分別由區塊C及區塊D向前對應至區塊C’及區塊D’。更 13 201119414 詳細地說，原始晝面Fn中的區塊c w中的區塊c，及縣畫面㈣中2=應至原始晝面 晝面Fn·2的區塊d，。最後’亦採用區塊:= 值（Sum of Absolute Difference )的方次、、異〜 C，的差異值_及區塊D和區塊D:的 樣地，再利用第二參考移動向量 塊D向後對應至晝面Fn+1的區塊 由&塊匚及區 D”，並計算產生區塊C和區塊c”的差里的區塊 D和區塊D”的差異值SAD8。 ，、細及區塊 最後’依據差異值SAD1〜SAD8同時 於影像覆蓋區或影像顯露區，以及區/龙〇位 量MVO為第-參考移動向量聰或第 請2。更詳細地說’當差編顧和差異值== 於一臨界值加時，區塊〇位於影像覆蓋區 = 動向量MV0為第一參考移動向量Μνι;當差里= 和差異值議都小於—臨界值TH1時，區塊〇、= 顯露區，且代表性移動向量MVQ為第 K當差異值SAD5和顧值⑽㈣持^界^ ^ 時，區塊〇位於影像覆蓋區，且代表性移動向量 第二參考義向量购；以及#差異值SAm和差= SAD8都小於一臨界值TH1 B主rs· ,λ, γλ ' 曰^ “ / 區塊0位於影像顯露區， 且代表性移動向lMV〇為第二參考移動向量贿2， 臨界值TH1為-預設值。請參照第仏〜仍圖之例了 ，例中第-參考移動向量贿為背景 (background motion vector) ^ ^ # ^ 201119414 為月景移動向量（foregroun(j m〇ti〇I1 vector )。另外，由第 4B圖可見，差異值SAD1為計算區塊A與區塊A，差異程 度，區塊A係位於背景區域，而區塊A，係位於前景區域， 戶^以差異值SAD1會很大；差異值SAD2為計算區塊B與 品鬼差異程度，區塊B係位於背景區域，而區塊b，係 • 位於前景區域，所以差異值SAD2亦會很大；差異值SAD3 • 區塊A與區塊A”差異程度，區塊a和區塊A”均位 ^月厅、區域’且由背景移動向量所對應，所以差異值SAD] _ :很小，差異值SAD4為計算區塊B與區塊B”差異程度， 區❹和區塊B”均位於背景區域，且由背景移動向量^對 …'斤以差異值SAD4會报小。同樣地，請參照第4D圖， 亦可知本範例中之差異值嶋5和SAD7很大，而差異值 SAD6和SAD8很小。依據前述之判定規則，本範例中差 =值SAD3和SAD4這—組同時小於―預纽，所以區塊 位於影像_區，且代紐移動向量Μν〇為第一參考 移動向量MV1。 ^The operation is coming. First, the semantic vector generation module (10) "the block comparison calculus calculates the motion vector of the miscellaneous block, and the calculated motion vector and the complex number of its neighboring blocks move to 1 ' to calculate the _ degree of the crane vector. In order to calculate a numerical value, in the embodiment, the numerical curve is a curve formed along the degree of confusion of a certain interpolation block vector, and the chaotic process and the motion vector variance, that is, , upwards represent a plurality of values of different moving vectors, and refer to the μ map, which represents the value of the degree of variation of the motion vector (4)^= 201119414 The patch block and its plurality of neighboring blocks (as shown in Figure 3a) 5x5 block range) According to the block motion comparison algorithm, the complex motion vector tool 疋 MVqo and MV·2·2~MV22 ' can calculate a motion vector variance value MV_VAR according to the motion vectors. It is calculated by taking the absolute value of the largest horizontal component of the motion vectors minus the minimum horizontal component, plus the largest vertical component of the motion vectors minus the minimum vertical component. For the value, MV_VAR can be expressed by the following equation: MV_VAR= | MAX(MVx)-MIN(MVx) | + I MAX(MVy)-MIN(MVy) | Equation (1) where MVX and MVy represent levels, respectively Component (x-axis component) and vertical component (y-axis component). It should be noted that the block range of 5x5 is not a limitation of the present invention, and it can also be implemented by using NxN or a block range of ΝχΜ, where The parameters Ν and Μ are both positive integers and Ν is not equal to Μ; in addition, the way to calculate the motion vector variogram MV_VAR can also be implemented by using equation (2) or equation (3): MV_VAR= I MAX(MVx)-MIN (MVx) | + I MAX(MVy)-MIN(MVy) | +SAD Equation (2) MV_VAR=ax{ | MAX(MVx)-MIN(MVx) | + I MAX(MVy)-MIN(MVy) | }+y3xSAD Equation (3) where the value SAD is the difference between the blocks of the first interpolation block according to the block comparison algorithm, and the parameters a and 0 are weighted parameters; It is within the scope of the invention to calculate any implementation variation of the value representative of the degree of variation of the motion vector. From the above, according to one of the equations (丨), equation (2) or equation (3), the motion vector The generation module 210 is different for each one The interpolation block of 201119414 is calculated, so that the first numerical curve cv can be obtained as shown in Fig. 3B. Fig. 3B is a first numerical curve CV including a plurality of moving vector variation values according to the present invention. «MB(K) is the first interpolation block, and in the 3B figure, the motion vector generation module 21 determines the two reference motion vectors MV: l, MV2 of the interpolation block mb(8), first, with the background Or the direction in which the foreground image moves (for example, the horizontal direction), the motion vector generation module 2(7)' is spatially extended by the plurality of sides of the first interpolation block MB00 (for example, the MB 〇 0 level is six on each side of the same The size of the block, mb10~MB6〇 and ΜΒ·1()~MB_6〇) the corresponding motion vector variation value, and take a maximum value within the range of the corresponding motion vector variation (for example, Figure 3B VARm ax )' and find the block corresponding to the minimum value of the left and right sides of the maximum value VARmax within the range of the corresponding motion vector variation values. For example, the block Na-4〇 can be found. MB5〇, and the left and right blocks MB-4〇 and MBs. The motion vector calculated by the block alignment algorithm, respectively, is the two reference motion vectors MV1, • MV2' of the first interpolation block MB00, in other words, the 'interpolation block MB'. . The first reference movement to fMV1 corresponds to the minimum value VARmin located to the left of the maximum value vARmax of the 帛-value series CV, and the second reference movement vector guesses the 2 system's position to the maximum value VARmax of the first-valued curve CV The minimum value VARmin' on the right side; the first and second reference motion vectors should correspond to the background motion vector (backgr〇und - _〇Γ), and the other corresponds to the foreground motion vector ( The foreground motion vector), because the motion vector variability around the interpolation block belonging to the image coverage area or the image exposure area will be quite large, and the left and right images of the minimum motion vector variation value on the 201119414 side correspond to the image. A block, which corresponds to a foreground shift or a background shift, depending on whether it is located in the image coverage area or image, or the dew area. Therefore, if the first interpolation block mb(10) is located in one of the image coverage area and the image display area, one of the first and second motion vectors is compliant, and one of the bodies corresponds to the background movement direction, and the other - Corresponds to the foreground motion vector. By the above operation, the moving orientation generating module 210 can calculate the corresponding first and second reference motion vectors 丨 and _2 for each of the interpolation blocks π. Next, please refer to 帛4A, 4B, 4C, and 4D diagrams as a schematic diagram of the first embodiment in which a mask block is located in the image overlay or image display area and its representative motion vector. The fuzzy block determining module 221 determines whether the block is located in the blurred area according to the difference between the first and second McCaw moving vectors MV1 and MV2 of the block. When the first and second reference motion vector reads of the block are similar to MV2, then the block is not located in the blurred area. When the first and second reference motion vectors of the block are different from Mv2, Then the block is in the blurred area. The difference value "乂一^^汗 is calculated by subtracting the absolute value of the horizontal component of the reference motion vector MV1 minus the horizontal component of the second reference motion vector MV2, plus the vertical component of the first reference motion vector MV1 minus the first The absolute value of the reference motion vector MV2 is 1 and the absolute value of the MVjDiff can be expressed by the following equation: MV_Diff= | MV1x-MV2x I + I MVly-MV2y | Equation (4) The above equation (4) is only The present invention - a preferred embodiment, is not a limitation of the present invention. Next, the overlay/exposure block determination and motion vector decision module 222 advances the coverage/exposure block determination for the block located in the blurred region and the generation 12 _ I _ 201119414 The eigenmotion vector is determined. In Fig. 4A, the interpolation plane FI is interpolated by the original pictures Fn-Ι and Fn to generate 'where the blocks in the region R3 and the region R6 are located in the blurred region. The following is an example of the region R3. The implementation of this embodiment is described in detail. A block to be determined is located in the region R3, and the block is obscured by the foreground object in the original face Fn-Ι, and does not appear in the original face Fn-1, but there is Now in the original face Fn (meaning area R) 3 is located in the image display: dew area, and the two motion vectors of block 0 are the first and second reference shifts respectively: motion vectors MV1 and MV2. First, referring to FIG. 4A, the second reference motion vector MV2 is used by the block. Ο corresponds to the block B of the original picture Fn-Ι and the block A of the original picture Fn; then, referring to FIG. 4B, the block A and the block B are respectively forwarded to the block by using the first reference motion vector MV1 Block A' and block B'. In more detail, the block A in the original face Fn corresponds to the block A' in the original face Fn-Ι and the block in the original face Fn-Ι Block B corresponds to the block B' of the original face Fn-2. Finally, the preferred embodiment calculates the block A and the block A' by means of the block comparison Sum of Absolute Difference. The difference value SADI and the difference value SAD2 of the block B and the block B'. Similarly, the first reference motion vector MV1 is reused from the block A and the block B, respectively, to the block A of the face Fii+1. And the block B' ' of the face Fn and calculate the difference value SAD3 of the block 'A and block A' and the difference value SAD4 of the block B and the block B". Then, referring to FIG. 4C, the block C corresponding to the original face Fn and the block D of the original face Fn-Ι are used by the first reference motion vector MV1; then, referring to the 4D figure, the second is utilized. The reference motion vector MV2 is forward-directed by the block C and the block D to the block C' and the block D', respectively. Further 13 201119414 In detail, the block c in the block c w in the original face Fn, and 2 in the county picture (4) = the block d which should be to the original face Fn·2. Finally, the block: the value of the Sum of Absolute Difference, the difference value of the different C, and the sample of the block D and the block D: are also used, and the second reference moving vector block D is used. The block corresponding to the face Fn+1 is backed by the & block and the zone D", and the difference value SAD8 of the block D and the block D" in the difference between the block C and the block c" is calculated. , fine and block finally 'based on the difference value SAD1 ~ SAD8 at the same time in the image coverage area or image display area, and the area / dragon position quantity MVO is the first - reference movement vector Cong or the second. More in detail 'when the difference Gu and difference value == When a critical value is added, the block 〇 is located in the image coverage area = the motion vector MV0 is the first reference motion vector Μνι; when the difference = = and the difference value are less than the - threshold TH1, the block 〇 , = the exposed area, and the representative motion vector MVQ is the Kth when the difference value SAD5 and the Gu value (10) (4) hold the ^^^^, the block 〇 is located in the image coverage area, and the representative motion vector second reference eigenvector is purchased; #差值SAm and difference=SAD8 are both smaller than a critical value TH1 B main rs· , λ, γλ ' 曰 ^ " / block 0 is located in the image display Region and a second reference motion vector representative mobile bribe 2 to lMV〇 is, the threshold TH1 is - a preset value. Please refer to the example of the second to the still figure. In the example, the reference motion vector is the background motion vector ^ ^ # ^ 201119414 is the moonscape vector (foregroun(jm〇ti〇I1 vector ). In addition, by It can be seen from Fig. 4B that the difference value SAD1 is the calculation block A and block A, the degree of difference, the block A is located in the background area, and the block A is located in the foreground area, and the difference value SAD1 of the household is large; The value SAD2 is the difference between the calculation block B and the character ghost, the block B is located in the background area, and the block b, the system is located in the foreground area, so the difference value SAD2 will also be large; the difference value SAD3 • block A and area Block A" difference degree, block a and block A" are all located in the moon hall, area ' and correspond to the background motion vector, so the difference value SAD] _ : is small, the difference value SAD4 is the calculation block B and area Block B" difference degree, zone 区 and block B" are located in the background area, and the difference value SAD4 will be reported by the background movement vector ^.... Similarly, please refer to the 4D figure, also in this example. The difference values 嶋5 and SAD7 are large, and the difference values SAD6 and SAD8 are small. Given rule, in this example the difference value = SAD3 and SAD4 this - at the same time less than the set - Pre New York, so the block is located in the image area _, New Generation and a first reference motion vector Μν〇 motion vector MV1 ^.
另外’睛參照第5Α、5Β、5C及5D圖係為決定一| 糊區塊位於影像覆蓋彡像顯親及其代表性移動向^ 之第二實施例示意圖。模糊區塊判定模組221係依據一运 塊之第-、第二參考移動向量聰與體的差異性決另 1塊位於影像覆蓋區或f彡軸露區。⑽地，當該區勒 的第-、第一|考移動向量咖與题相似時，則該逼 ，不位於模糊區域；‘然當該區塊的第―、第二參考移誠 里MV1與MV2差異大時，則該區塊位於模糊區域。且差 異值MV_Diff輯算枝亦與f —實施例相同。 15 201119414 接著，覆蓋/顯露區塊判定及移動向量決定模組222進 ^對位於拉糊區域之區塊進行覆盖/顯露區塊判定及代 表性移動向量決定。在第5A圖中，插補畫面FIn由原始 晝面Fn-：^aFn所插補產生，其中區域R3和區域中的 區塊位於模糊區域。以下以區域R3為例，詳細說明本實 施例作法。一待決定區塊Ο位於區域R3中，區塊〇於原 始晝面Fn-Ι中被前景物體所遮蔽，未出現在原始晝面化-工 中’但有出在原始晝面Fn中（意即區域R3位於影像顯 ==區塊。的兩移動向量分別為第一、第二參考移 動向$ MV1與MV2。首先，參照第5A圖，利用第一參 考移動向量MV1由區塊〇對應至原始晝面如的區 ^ 原始晝面Fn-Ι的區塊F ;接著，參照第5B圖，利用第二 向量，MV2由區塊E及區塊F向前分別對應至ί Α區塊F。更詳細地說，原始晝面Fn中 前對應至插補晝面FIn中的區塊£，及原始畫面= 塊F對應至插補書面打 ' ⑹旦㈤e 塊？，再比較區塊E，的移 ^置聰和區塊F，的移動向量聊 ，移動向量MV2。同樣地，再利用 .MV2分別由區塊E及 "多動向里 ，再咖塊應為塊E，，及區塊 y 移動向i MVE和區塊F”的移動向 里卿是否相似於第二參考移動向量㈣。移動白 然後，參照第5C圖，利用第 分麻塊⑽塊Η向前對二參== 201119414 =也說，原始晝面Fn中的區塊G向前對應至插補晝面 η的區塊G’及原始晝面㈤中的區塊^對應至插補晝 2 _的區塊Η’ ’再比較區塊G’的移動向量Μ·和區 2 的移動向量MVH’衫相似於第—參考移動向量 。同樣地，再利用第—參考移動向量讀分別由區 • ^區塊㈣後對應至區塊G”及區塊H，，，再比較區塊 • =動向量贿，和區塊H”的移動向量麵，，是否相 似於第一參考移動向量MV1。 • 取後，依據上述移動向量MVE，、MVF，、MVE”、 ==着、着、職，，及_，，來同時決定區塊〇 2 =像覆蓋區或影像顯露區，以及區塊〇的代表性移動 MV里2 為卜參考飾向量簡或第二參考移動向量 —#細地說’當移動向量MVE，和移動向量MVF， 目嫌第二參考移動向量购時，區塊〇位於影像覆 Mi.’H表性移動向量MV〇為第二參考移動向量 -失二f VE”轉麟4 MVF，，都與相似於第 量娜2時’區塊0位於影像顯露區，且代 二夕向置MVO為第二參考移動向 量MV2 ; ^移動向 里MVG和移動向量mvh，In addition, the reference to the fifth, fifth, fifth, and fifth images is a schematic diagram of the second embodiment in which the paste block is located in the image overlay image and its representative moving direction. The fuzzy block determination module 221 is based on the difference between the first and second reference motion vectors of the one block and the other is located in the image coverage area or the f彡 axis. (10) Ground, when the first-, first-test movement vector coffee of the area is similar to the question, then the force is not in the fuzzy area; 'When the first and second reference of the block are moved to the MV1 and When the MV2 difference is large, the block is located in the blurred area. And the difference value MV_Diff is also the same as the f-embodiment. 15 201119414 Next, the overlay/exposure block decision and motion vector decision module 222 performs a coverage/exposure block decision and a representative motion vector decision on the block located in the pad area. In Fig. 5A, the interpolation picture FIn is interpolated by the original plane Fn-: ^aFn, wherein the area R3 and the block in the area are located in the blurred area. The practice of this embodiment will be described in detail below by taking the region R3 as an example. Once the block is determined to be located in the area R3, the block is obscured by the foreground object in the original face Fn-Ι, and does not appear in the original facet-worker but is in the original face Fn (meaning That is, the two motion vectors of the region R3 are located in the image display == block, and the first and second reference movements are respectively toward the MV1 and MV2. First, referring to the fifth reference diagram, the first reference motion vector MV1 is used to correspond to the block by the first reference motion vector MV1. The original facet area ^ the original face Fn-Ι block F; then, referring to FIG. 5B, using the second vector, MV2 is forwarded by block E and block F to the ί block F, respectively. In more detail, the original face Fn corresponds to the block in the interpolation face FIn, and the original picture = block F corresponds to the interpolated written hit '(6) denier (five) e block?, and then compares block E, Move the set and the block F, move the vector chat, move the vector MV2. Similarly, reuse the .MV2 by the block E and " multi-directional, then the block should be the block E, and the block y Move to i MVE and block F" to move to Liqing whether it is similar to the second reference movement vector (4). Move white then, refer to Figure 5C, use the The hemp block (10) block Η forward pair two parameters == 201119414 = also said that the block G in the original face Fn corresponds to the block G' of the interpolation face η and the block in the original face (5) ^ corresponds to the interpolation 昼 2 _ block Η ' ' re-comparison block G ' the movement vector Μ · and the area 2 movement vector MVH ' jersey similar to the first - reference movement vector. Similarly, reuse the first reference The motion vector reads are respectively corresponding to the block G" and the block H, and then the block vector = = dynamic vector bribe, and the moving vector plane of the block H", respectively, is similar to the A reference motion vector MV1. • After taking, according to the above motion vectors MVE, MVF, MVE", == 着, 着,职,, and _, to determine the block 〇2 = like coverage area or image revealing The region, as well as the representative mobile MV of the block 2 2 is the reference vector or the second reference moving vector - # 细 细 'When the moving vector MVE, and the moving vector MVF, the second reference mobile vector purchase time , block 〇 is located in the image overlay Mi. 'H eigenistic motion vector MV 〇 as the second reference motion vector - lost two f VE" to Lin 4 MVF They are similar to the amount of Na 2 'exposed image region located block 0, and the generation of two opposite MVO Xi is moved to the second reference amount MV2; ^ mvh moved inwardly and MVG motion vector,
Μ ^ 〜像覆起，且代表性移動向量MVO 二彡矛夕動向里MV1;以及當移動向量MVG”和移動 似於第一參考移動向量題時，區= 動且代_多動向量mv〇為第一參考移 勤向里MV1。請參昭筮s Δ 參考移動向量，本範例中第一 為月豕私動向量（background motion 17 201119414 vector )，而第二參考移動向量MV2為前景移動向量 (foreground motion vector)。另外，由第 5B 圖可見，區塊 E’係位於背景區域，而區塊F’係位於前景區域，所以移動 向量MVE’與第二參考移動向量MV2不相似，但移動向量 MVF’與第二參考移動向量MV2相似；區塊E”係位於背予_ 區域，而區塊F”係位於前景區域，所以移動向量MVE”盘 第二參考移動向量MV2不相似，但移動向量MVF”與第-參考移動向量MV2相似。同樣地，請參照第5D圖，亦可 知本範例中之移動向量MVG，和MVH，與第一參考移動向 量MV1均不相似’而移動向量MVG，> MVH”與第一泉考 移動向量MV1均相似。依據前述之判定規則，本範例中移 動向量MVG”和MVH”與第一參考移動向量MV1的差異值 小於一預定值，所以區塊〇位於影像顯露區，且代表性 動向量MV0為第一參考移動向量MV1。 夕 最後，晝面插補模組230依據該代表性移動向 區塊位於影像覆蓋區或影像顯露區之結果， : 的插補晝面影像。詳細地說，當區塊〇位於影像顯=塊 ΐ代f移動向量_為第一參考移動向量MV^以 弟一參考移動向量職向後對應至縣晝面&的區換 補區塊0，·當㉛切位於影細 _為第二參考移動向量㈣，則以第m夕Γ MV2向後鉗弟一參考移動向量 门傻對應至原始晝面Fn的區塊填補 里 0位於影像覆蓋區，且代表性 ，*區塊 移動向量MV1，則以第一夫告获丨里曰0為第—參考 原始晝面Fn-Ι的區塊填補^ ^_里MV1向前對應至 L鬼填補£塊〇，“塊〇位於影像覆蓋 18 201119414 區^代表性飾向量MV〇為第二參考軸向量雨2， 則以第—參考移動向f MV2向前對應至原始晝面㈤的 區塊填補區塊〇。在第4A圖或第5A圖的範例中，依據模 1區塊處理模組22G之決定，區塊〇位於影像顯露區，且 代表性移動向量MV〇為第一參考移動向 量MV1，則晝面 ： 插補她23Q會以第—參考移動向量MV1向後對應至原始 晝面Fn的區塊填補區塊〇。 * 此外，若一插補區塊實際屬於模糊區域（影像覆蓋區 • 或影像顯露區）’則畫面插補模組230依據模糊區塊處理模 組220所決定之區域及代表性移動向量，對該區塊來產生 正麵影像’因而可提升影像品質，其中該模糊區塊的代 表f1 生移動向里係對應於背景移動向量；而若該插補區塊實 際並非屬於模祕域（影像覆蓋區或影像顯露區），則晝面 插補模組230依據區塊比對方法對該插補區塊之產生的移 動向量來產生正確的影像。最終，畫面插補模組23〇產生 插補晝面Fin。 • 杯照第6 ® ’㈣本發明另-實施例之影像處理方 ' 法及其相_作步驟的流程。财法包括： • 步驟謝：產生該插補晝面（亦即第-插補晝面）中之一 插補區塊的第一參考移動向量MV1與第二參考移動向量 脚2。更詳細地說’首先，計算該插塊之相關區塊的 移動向量，並依據每一個相關區塊所計算出的移動向量與 其複數個鄰近區塊的複數個移動向量，來計算該些相關區 塊移動向量的混亂程度，以算出一數值曲線。接著，依據 該數值曲線蚊出該插補區塊兩參考移動向量 MV1、 201119414 2。詳細實施方式請參考前述說明書關於第从及犯 的對應說明，不另贅述。 MV2 步驟603 :依據該區塊之第一參考移動向量mvi盘第二參 2移動向量MV2來決賴區塊位料彡像覆舰或影像顯 路區’及並決定該區塊之-代表性移動向量。苴中，一較 佳實施方法先依據該區塊之第一、第二參考移動向量讀 與MV2的差異性決定該區塊是否位於模糊區域。接著，對 位於模糊區域之區塊先以第一參考移動向量Mv卜對該插 補晝面之前-張原始晝面及後一張原始晝面中找到對岸之 ^-區塊及第二區塊，再以第二參考移動向量應，從該 I-張原始晝面的第-區塊及該後—張原始晝面第二區塊 分別向其前-張原始畫面中找到對應之第三區塊及第四區 塊，及向其後一張原始晝面中找到對應之第五區塊及第六 區塊，並據以计异其對應區塊之相似性，以判定該插補區 塊位於復蓋區或顯路區判定及其代表性移動向量，詳細方 法請參考前述說明書關於第4A〜4D圖的對應說明，不另 贅述。亦可以另一較佳實施方法，對位於模糊區域之區塊 先以第一參考移動向量MV1，對該插補畫面之前一張原始 晝面及後一張原始畫面中找到對應之第一區塊及第二區 塊，再以第二參考移動向量MV2，從該前一張原始畫面的 第一區塊及該後一張原始晝面第二區塊分別向其前一張插 補畫面中找到對應之第三區塊的移動向量及第四區塊的移 動向虽’及向其後一張插補晝面中找到對應之第五區塊的 移動向3：及第六區塊的移動向量’並將第五區塊的移動向 量及第六區塊的移動向量分別與第二參考移動向量 20 201119414 =比，叫賴插娜塊⑽覆蓋 其代表性移動向量，詳細方法請參考前述說定及 〜5D圖的對應說明，不另贅述ά 田闕於第5Α =605 ··依據該移動向量及該區塊位於 f顯露區之結果，來產生該區塊的插補畫面‘ f或影 祝，當該區塊位於影像顯露區 =。汗細地 影像覆蓋區，===該!塊;當區塊位於 始晝面的區塊填補區塊。、α里向别對應至前一張原 —以上為本發明影像處理方法的步驟m =到相同的結果’並不需要_ ;示:二 3步:順序來進行，且“圖所示之步驟不一 進仃、，亦即其他步驟亦可插入其中。 疋要物 專利!僅為本發明之較佳實施例，凡依本發明申, r圍所做之均等變化與修飾，皆應屬本發明二； 【圖式簡單說明】 第1 第2尾比對肩异法的操作示意圖。 ，異程度之數值的範二產生‘組计异代表移動 θ為匕3有多個移動向量變異值之數值曲線的範例 21 201119414 不意圖。 第4A〜4D Bl A楚ο η 固与弟2圖所示之模糊區塊處理模組的第一實 施例之示意圖。 第5A〜5D圖為第2 施例之示意圖。 圖所示之模糊區塊處理模組的第二實 第圖為本發明影像處理方法之流程圖。 【主要元件符號說明】 本案圖式中所包含之各元件列示如下： 1〇影像處理裝置 100儲存單元 200移動插補單元 210移動向量產生模組 220模糊區塊處理模組 221模糊區塊判定模組 222覆蓋/顯露區塊判定及移動向量決定模組 230晝面插補模組 22Μ ^ ~ like the overlay, and the representative motion vector MVO 彡 动 MV MV1; and when the motion vector MVG" and the movement is like the first reference motion vector problem, the region = the dynamic and the _ multi-motion vector mv〇 For the first reference, shift to the inner MV1. Please refer to the 筮 Δ Δ reference motion vector. In this example, the first is the lunar motion vector (background motion 17 201119414 vector ), and the second reference motion vector MV2 is the foreground motion vector. (foreground motion vector). In addition, as seen from Fig. 5B, the block E' is located in the background area, and the block F' is located in the foreground area, so the motion vector MVE' is not similar to the second reference motion vector MV2, but moves The vector MVF' is similar to the second reference motion vector MV2; the block E" is located in the back _ region, and the block F" is located in the foreground region, so the motion vector MVE" disk second reference motion vector MV2 is not similar, but moves The vector MVF" is similar to the first reference motion vector MV2. Similarly, referring to the 5D picture, it can be seen that the motion vector MVG in this example, and MVH, are not similar to the first reference motion vector MV1, and the motion vector MVG, ≫ MVH" is similar to the first spring motion vector MV1. According to the foregoing determination rule, the difference value between the motion vectors MVG" and MVH" and the first reference motion vector MV1 in this example is less than a predetermined value, so the block 〇 is located in the image exposure area, and the representative motion vector MV0 is the first reference. Move the vector MV1. Finally, the facet interpolation module 230 interpolates the face image according to the result of the representative moving direction block located in the image coverage area or the image display area. In detail, when the block 〇 is located in the image display = block generation f movement vector _ is the first reference movement vector MV ^ to the reference to the mobile vector position after the reference to the county face & area replacement block 0, · When 31 is located in the shadow _ is the second reference motion vector (four), then the first m Γ MV2 backwards the tongs a reference movement vector door silly corresponding to the original face Fn block fill 0 is located in the image coverage area, and Representative, *block movement vector MV1, then the first answer is 丨里曰0 as the first - reference to the original face Fn-Ι block fill ^ ^ _ MV1 forward corresponds to L ghost fill block 〇 , "block 〇 is located in image coverage 18 201119414 area ^ representative decoration vector MV 〇 is the second reference axis vector rain 2, then the first reference movement to f MV2 forward corresponding to the original surface (five) block filling block In the example of FIG. 4A or FIG. 5A, according to the decision of the modulo 1 block processing module 22G, the block 〇 is located in the image exposure area, and the representative motion vector MV 〇 is the first reference motion vector MV1, then昼: Interpolating her 23Q will correspond to the block of the original face Fn with the first reference motion vector MV1 In addition, if an interpolation block actually belongs to a blurred area (image coverage area or image display area), the picture interpolation module 230 determines the area and representative according to the fuzzy block processing module 220. Sexual motion vector, which produces a frontal image for the block. Thus, the image quality can be improved, wherein the representative of the fuzzy block moves inwardly corresponding to the background motion vector; and if the interpolation block does not actually belong to the mode In the secret domain (image coverage area or image display area), the face interpolation module 230 generates a correct image according to the block vector to the motion vector generated by the interpolation block. Finally, the picture interpolation module 23〇Generate the interpolation Fin. • Cup photo 6® '(4) The image processing method of the other embodiment of the invention and the process of the phase thereof. The financial method includes: • Step Xie: Generate the interpolation One of the facets (ie, the first-interpolation facets) interpolates the first reference motion vector MV1 of the block and the second reference motion vector pin 2. In more detail, first, the relevant block of the block is calculated. Moving vector and based on each Calculating a chaotic degree of the motion vector of the relevant block by calculating a motion vector calculated by the relevant block and a plurality of motion vectors of the plurality of neighboring blocks to calculate a numerical curve. Then, according to the numerical curve, the mosquito is inserted Complement block two reference motion vectors MV1, 201119414 2. For detailed implementation, please refer to the foregoing description of the corresponding description of the slave and the offense, no further details. MV2 Step 603: According to the first reference mobile vector mvi disk of the block Reference 2 moves the vector MV2 to rely on the block bit map or image display area 'and determines the representative moving vector of the block. In the middle, a preferred implementation method first depends on the block 1. The difference between the second reference motion vector read and MV2 determines whether the block is located in the blurred area. Then, the block located in the blurred area first finds the opposite block and the second block in the front of the original face and the next original facet with the first reference motion vector Mv. And the second reference motion vector should be used to find the corresponding third region from the first block of the original slice and the second block of the subsequent original slice And the fifth block and the sixth block are found in the next block and the next original block, and the similarity of the corresponding block is calculated to determine the interpolation block. It is located in the coverage area or the display area and its representative motion vector. For details, please refer to the corresponding description of the above description on the 4A to 4D drawings, and will not be described again. In another preferred implementation method, the first reference motion vector MV1 is first used in the block located in the blurred area, and the corresponding first block is found in the original original picture and the next original picture in the previous picture. And the second block, and then using the second reference motion vector MV2, respectively, from the first block of the previous original picture and the second block of the next original picture to the previous interpolation picture Corresponding to the movement vector of the third block and the movement of the fourth block, and the movement vector of the third block and the sixth block are found in the subsequent interpolation block. 'And the motion vector of the fifth block and the motion vector of the sixth block are compared with the second reference motion vector 20 201119414 respectively, and the Lai Na Na block (10) covers its representative motion vector. For details, please refer to the foregoing description. And the corresponding description of the ~5D map, no further description ά Tian Yu at the 5th Α = 605 · According to the movement vector and the result of the block located in the f exposure area, to generate the interpolation picture of the block 'f or shadow When the block is located in the image exposure area =. Sweat fine image coverage area, === this! block; when the block is located at the beginning of the block fill block. , α corresponds to the previous original - the above is the step of the image processing method of the present invention m = to the same result ' does not need _; shows: two 3 steps: the order is carried out, and the steps shown in the figure Other steps may also be inserted into it. The patents are only preferred embodiments of the present invention, and the equivalent changes and modifications made by the invention according to the present invention should belong to the present invention. Invention 2; [Simple diagram of the diagram] Schematic diagram of the operation of the first and second tails of the shoulder-to-shoulder method. The value of the difference of the value of the second generation of the group of the two groups represents the movement θ is 匕3 has a plurality of motion vector variation values Example of a numerical curve 21 201119414 It is not intended. 4A to 4D bl A diagram of the first embodiment of the fuzzy block processing module shown in Fig. 5A to 5D. The second actual figure of the fuzzy block processing module shown in the figure is a flow chart of the image processing method of the present invention. [Description of main component symbols] The components included in the drawing of the present invention are listed as follows: The image processing device 100 storage unit 200 moves the interpolation unit 210 to move Amount of blur generation module 220 blocks the fuzzy processing module 221 determines that the block module 222 cover / expose block and the motion vector determination module 230 determines surface interpolation module 22 day